Pascale Fung

Cofounder and CRIO of AMI Labs. Chair Professor of ECE, HKUST. Fellow of AAAI, ACL, IEEE, ISCA.

[Pascale Fung @pascalefung]

@shayekhbinislam @mareksuppa I will publish my writing of the keynote speech. The prompts are basically turn this into 45 slides and then reprompt individual slides, add or delete slides, to get what I want. The videos were a hoot to generate. Claude included music to the videos which I muted.

[lifcc @mylifcc]

为什么现在的顶尖大模型说话越来越“拗口、难懂”了？
NYU 和 UT Austin 刚刚发了一篇非常扎心的心理语言学论文，直接扒开了现代大模型“不好好说人话”的底牌。
研究人员把 Claude Sonnet 5 / Opus 5.5、GPT-5.6、GLM-5、Kimi K2.5 等 9 款前沿模型拉出来做全方位句法解剖，发现了 3 个反直觉真相：
1️⃣ 顶尖 AI 正在滥用 6 种“高认知负荷”句式
相比真实人类写作和同义回译文本，前沿模型极度偏爱：省略 that 的宾语从句、多重名词硬堆叠（如 reader-patience budget）、名词强行作动词（如 calendar it / footnote it）、无生命主语配动作动词，以及超长句法依存距离。这些特征在心理语言学里全是消耗人类短期记忆的“阅读杀手”。
2️⃣ 根源在于“人类大脑”与“推理后训练”的错位
人类语言遵循“均匀信息密度（UID）”——为了照顾听众有限的工作记忆，我们会本能地加入过渡词把信息摊平。而现代大模型经过大量数学、代码和推理（RLVR）后训练，只追求在有限窗口内最大化“信息密度”。
3️⃣ 一个简单的长度惩罚，就能复刻这种“反人类文风”
团队在微调实验中加入长度惩罚（Length Penalty），逼模型少用 Token，结果模型立刻学会了砍掉连接词、堆叠名词，瞬间变成那股熟悉的“高密度 AI 味”。
当模型为了解题效率疯狂压缩 Token 时，你觉得未来的 AI 应该区分“高密度推理模式”和“人类友好沟通模式”吗？

[Arnaud Bertrand @RnaudBertrand]

The ugly truth about France and some of the other European countries currently in turmoil - and which none of the politicians want to admit because it'd be political suicide - is that you can't solve the problem by going after immigrants (as the right asserts) or the rich (as the left asserts), but you could solve it by going after the old.

One extraordinary statistic for you: people over 60 in France, a quarter of the population, own 60% of France's wealth, non-financial and financial (https://t.co/IdJ5X7OO12).

And here is another crazy statistic:
that same quarter of the population, those above 60, receive nearly two-thirds of all social spending by the French government (https://t.co/deNRNevIwq) - which is famously among the highest in the world (France's government spending is an insane 57.3% of GDP).

Last statistic: they make up about 40% of voters because they're a third of registered voters (https://t.co/zTa4RnVOmT), and they turn out far more than anyone else (https://t.co/34V2xzs9jH).

Put our 3 statistics together and there's almost a bit of an apartheid situation here: the old own most of the country's wealth, collect most of the social spending and have the political power to keep it that way.

Which makes them a de-facto rentier class, living off assets and transfers paid for by younger generations.

And, of course, at the risk of stating a truism, the future is built by the young. Yet France keeps effectively taking money away from them to give it to those whose working lives are already behind them - and need the money far less since they already own the country's wealth.

What about inheritance, I hear you ask? That's the funniest/saddest aspect: given that life expectancy has kept rising, the French now inherit a home at 54 on average. The most common case is someone in their 80s leaving it to heirs in their 50s, in effect the very old transferring wealth to the about-to-be old (https://t.co/d2RPAqLJGR).

You can add this to our earlier apartheid statement: on top of all the rest, even when the old finally let go of their wealth, it stays within their own "camp".

That's my personal situation, incidentally: I'm 42, my father is 71 and he is starting to look at ways to pass things on to me and my siblings, when our professional lives are already well settled, so it'll arrive when it changes things the least, which is pretty much the French system in a nutshell.

Now, I get that there'd be some unfairness in going after them: they paid into the system their entire working lives, playing by the rules of a social contract that told them if they paid up they'd collect at the end. Changing the rules now means breaking that promise, which is admittedly unfair.

But there's no magic wealth tree in any society: you can only take it where it is. That's the way most major societal transformations end up happening. They're often triggered by a catastrophic event (the so-called "four horsemen": mass-mobilisation war, revolution, state collapse, plague) BUT what really drives change is a transfer: wealth gets pried from the privileged class who've been sitting on it and handed to those who'll build what comes next.

So, in theory and in an ideal world, you can do without the catastrophic event and directly proceed with the transfer...

The fact that in France, and elsewhere in Europe, it's the youth protesting is probably no coincidence - even though they don't frame it as a generational fight. And even though, cruel irony, the left politicians many of them march behind promise to bring the retirement age back to 60 (https://t.co/WrJOhpWqYR), which would make the situation even worse.

Let's say they understand intuitively that the system is stacked against them, even if they haven't yet figured out how. And politicians are obviously doing what politicians do, meaning pandering to the people who actually vote, while pointing the young towards easier - and politically convenient - scapegoats.

To be clear I'm not saying it's the only problem in France, VERY far from it. I'm also not saying that all old people in France live a lavish lifestyle at the expense of younger generations: there are also plenty of very poor retirees who, if anything, would deserve more.

What I am saying is that it's probably the biggest elephant in the room: everyone can see it, nobody dares mention it - no politician, from any side, wants to run on cutting grandma's pension - and it makes every other problem harder to fix.

[Pascale Fung @pascalefung]

Personal stylist? This ought to be a low hanging fruit!

[Pascale Fung @pascalefung]

All the top chatbots, ChatGPT, Gemini, DeepSeek etc. have terrible fashion sense. No RLHF for fashion?

[Pascale Fung @pascalefung]

Anthropic has become a freaking cult! You want people to care about machine consciousness? PROVE there is one. Can’t prove it?! You are just a cult leader @DarioAmodei. You are damaging people’s minds with heresay.

[Pascale Fung @pascalefung]

AI safety is science, not just philosophy and certainly is not politics. Humanity is doomed if we turn every scientific subject into politics.

[Pascale Fung @pascalefung]

@PaulBonnet I can assure you that nobody needs my congratulations.

[Pascale Fung @pascalefung]

I am looking to hire PhD students, post-docs, and ML researchers in our Paris office interested in collaborating on fundamental research in AI model safety, not safety rules, not AI policy, not post hoc guardrails, not RLHF. We also have offices in Singapore, NYC, and Montreal. @amilabs #aisafety

[Pascale Fung @pascalefung]

The recording of my ICML 2026 keynote is online. I explain why real world problems can only be solved by world models not generative models in real-time.
https://t.co/CierSDvUxC

[Pascale Fung @pascalefung]

We can argue forever whether the term should be LLM “hallucination” or “confabulation” but language is a democratic thing. The global AI scientific community has adopted the term “hallucination” because it is easy to understand and remember and easier to pronounce. Only the anglophone elite promots the term “confabulation” and it doesn’t work because nobody else uses the word in daily contexts.

[Pascale Fung @pascalefung]

Jensen Huang promotes open-source LLMs for innovation. Great. The next step is democratizing compute. High GPU costs remain one of the biggest barriers to academic research and grassroots innovation. Imagine the impact if NVIDIA also expanded free or heavily subsidized compute for researchers and open-source projects. Open models need accessible compute to truly unlock innovation.

[Pascale Fung @pascalefung]

Jensen Huang promotes open-source LLMs for innovation. Great. The next step is democratizing compute. High GPU costs remain one of the biggest barriers to academic research and grassroots innovation. Imagine the impact if NVIDIA also expanded free or heavily subsidized compute for researchers and open-source projects. Open models need accessible compute to truly unlock innovation.

[Pascale Fung @pascalefung]

@deanwball Tell me you don’t know China without telling me you don’t know China.

[Pascale Fung @pascalefung]

Speaking from my American experience only here. Some people complain about large industry funded research. Don’t forget that almost all of paradigm changing research that led to modern AI came from either large industry labs - Bell Lab, IBM Labs, Xerox Parc, Microsoft, Google, GDM, FAIR, etc etc. or from the millitary defense funding. For modern AI, ideas that come from universities simply are not scabale to real impact in our area without industry research taking over. This was true before the era of LLMs. Many people in academic without industry experience also do not know what it takes to scale AI research to real production impact. So my (radical) view is that universities should work with industry if we want to have real impact OR focus on more theorectical and field research that does not require resources from either the private sector or the defense budget. More theoretical research does not need such funding. Linguistics certainly doesn’t. So doing research in language and language per se, not LLMs, not chatbots and not agents, does not need to be in any industry lab.

[DCI Resources, LLC @DCIResourcesLLC]

https://t.co/X8xAFyARMr

[Pascale Fung @pascalefung]

Giving my keynote at #icml in one hour https://t.co/CierSDwsna

[Pascale Fung @pascalefung]

Thrilled to announce the Wearable AI Workshop at ECCV 2026 🎉 that we're organizing with an awesome group of folks across Meta Reality Labs, AMI Labs, HKUST, Georgia Tech, UCF, and U. of Edinburgh. If you work on egocentric perception, multimodal assistants, or long-form video understanding, definitely consider participating. Details below 👇

🔬 Wearable AI Workshop @ ECCV 2026 | Malmö, Sweden
Join us to advance the science of real-time, multimodal, context-aware assistants — with access to the largest egocentric wearables dataset ever released, $21K in challenge prizes, and a venue to shape the research agenda alongside leading groups in the field.

🏆 Grand Challenge — $21K in Prizes
Compete on the largest egocentric wearables dataset ever released (5,000+ hours) across three tracks:
Proactive AI — Predict when and how to proactively engage users
Multi-Turn Conversations — Streaming answer prediction from egocentric video
Long Video Q&A — Temporal grounding and recall over 30+ minute videos

🎤 Confirmed Keynote Speakers
Kristen Grauman (UT Austin) — Professor, Computer Vision Group
Dima Damen (U of Bristol, Google DeepMind) — Professor of Computer Vision
Raffay Hamid (Meta Reality Labs) — Distinguished Engineer, GenAI

If you work on egocentric vision, long-form video understanding, conversational AI, or on-device inference — this is your venue.

📅 September 2026 · 📍 Malmö, Sweden
🔗 Workshop URL: https://t.co/XOfCNftuWq

#Meta #MetaRealityLabs #MetaAI #AMIlabs #HKUST #UTAustin #GeorgiaTech #UniversityOfEdinburgh #UniversityOfBristol #ECCV #ECCV2026 #WearableAI #EgocentricVision #ComputerVision #MultimodalAI #AIResearch

[Pascale Fung @pascalefung]

Machines do not have intents. Humans do. Machines do not have feelings. Humans do. Machines do not have psychology. Humans do. Machines can misbehave, but only humans can be evil. Stop virtual signaling and ethics washing and focus on real work on responsible AI, which starts with being responsible humans. #anthropic